Article Text

NAFLD and increased risk of cardiovascular disease: clinical associations, pathophysiological mechanisms and pharmacological implications
Free
1. Giovanni Targher1,
2. Christopher D Byrne2,
3. Herbert Tilg3
1. 1 Endocrinology and Metabolism, University of Verona Department of Medicine, Verona, Veneto, Italy
2. 2 Southampton National Institute for Health Research Biomedical Research Centre, University Hospital Southampton, Southampton, UK
3. 3 Department of Internal Medicine I, Gastroenterology, Hepatology, Endocrinology & Metabolism, Medical University of Innsbruck, Innsbruck, Tirol, Austria
1. Correspondence to Professor Giovanni Targher, Endocrinology and Metabolism, University of Verona Department of Medicine, Verona 37126, Italy; giovanni.targher{at}univr.it

## Abstract

Non-alcoholic fatty liver disease (NAFLD) is a public health problem, affecting up to a third of the world’s adult population. Several cohort studies have consistently documented that NAFLD (especially in its more advanced forms) is associated with a higher risk of all-cause mortality and that the leading causes of death among patients with NAFLD are cardiovascular diseases (CVDs), followed by extrahepatic malignancies and liver-related complications. A growing body of evidence also indicates that NAFLD is strongly associated with an increased risk of major CVD events and other cardiac complications (ie, cardiomyopathy, cardiac valvular calcification and cardiac arrhythmias), independently of traditional cardiovascular risk factors. This narrative review provides an overview of the literature on: (1) the evidence for an association between NAFLD and increased risk of cardiovascular, cardiac and arrhythmic complications, (2) the putative pathophysiological mechanisms linking NAFLD to CVD and other cardiac complications and (3) the current pharmacological treatments for NAFLD that might also benefit or adversely affect risk of CVD.

• nonalcoholic steatohepatitis
• cardiovascular disease
• cardiovascular complications

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

## Introduction

The 2017 Global Burden of Disease Study showed that there were 2.14 million liver-related deaths (2.06–2.30 million) that year, representing a 11.5% increase since 2012. Liver cancer and cirrhosis accounted for most of these deaths and, although chronic viral hepatitis remains the most common cause of liver-related death worldwide, these data show that non-alcoholic fatty liver disease (NAFLD) is the most rapidly growing contributor to liver-related mortality and morbidity.1 In 2016, it was estimated that in the USA, over 64 million people had NAFLD, with annual direct medical costs of about US$103 billion (US$1613 per patient), and in four European countries (France, Germany, Italy and UK), it was estimated that there were ∼52 million people with NAFLD with an annual cost of about €35 billion (from €354 to €1163 per patient). Costs of NAFLD were highest in patients aged 45–65 years and it was in this working age group where the economic costs of cardiovascular disease (CVD) were also much higher.2

CVDs, which include ischaemic heart disease and stroke, are the most common non-communicable diseases globally, responsible for an estimated 17.8 million deaths in 2017, of which more than three quarters were in low-income and middle-income countries.3 At the global scale, total deaths from CVD increased by nearly 21% between 2007 and 2017, and were greater for men than for women at most ages in 2017, except for ages≥85 years where there was the largest female-to-male ratio of CVD deaths.3

NAFLD occurs in at least 25%–30% of adults in high-income countries and in up to 70%–90% of individuals with obesity or type 2 diabetes mellitus (T2DM).4 NAFLD is also an important contributor to morbidity in other organs beyond the liver and, specifically, NAFLD is closely associated with an increased risk of developing extrahepatic complications, such as CVD, T2DM and chronic kidney disease, with fibrosis stage being the strongest disease-specific risk factor.5–8

This review article focuses on the rapidly expanding body of clinical evidence that supports a strong association between NAFLD and the risk of CVD, discusses the pathophysiological mechanisms that link these two conditions and summarises the pharmacological treatments for NAFLD that might also benefit or adversely affect risk of CVD.

### Risk of CVD and other cardiac complications

That NAFLD is associated with an increased risk of CVD is perhaps not surprising, given the close associations of NAFLD with cardiometabolic risk factors encapsulated by the metabolic syndrome, including abdominal obesity, atherogenic dyslipidaemia, hypertension and dysglycaemia.9 10 However, the nature and extent of the associations between NAFLD and CVD is not clear. Whether liver disease in NAFLD confers any additional CVD risk, or whether an increase in CVD risk in NAFLD is due to associated CVD risk factors, is uncertain. Elucidating whether liver disease in NAFLD contributes additional CVD risk is important, as it is plausible that treatment of liver disease may ameliorate risk of CVD, over and above treatment of NAFLD-associated risk factors.

Strong evidence links NAFLD with objectively assessed subclinical atherosclerosis (including also increased coronary artery calcium score) in adults and adolescents, as well as with an increased prevalence of clinically manifest CVD both in the general population and in different patient groups.11–13 Recently, in a large cohort of South Korean middle-aged individuals without pre-existing CVD, Lee et al also showed that imaging-defined NAFLD was independently associated with a higher risk of having non-calcified, ‘vulnerable’ coronary atherosclerotic plaques (as detected by coronary CT angiography), thereby highlighting an increased NAFLD-related CVD risk among these asymptomatic individuals.14 Several cohort studies have consistently documented that NAFLD is associated with a higher risk of all-cause mortality and that patients with NAFLD are more likely to experience a CVD-related death than a liver-related death.2 6 9 10 Using mortality data from the National Vital Statistics System multiple-cause mortality data in the USA, Paik et al recently confirmed that CVD was one of the most important causes of death among people with NAFLD.15 Several cohort studies have also shown that NAFLD (defined radiologically or histologically) is predictive of incident CVD events. Many of these studies were also included in a comprehensive meta-analysis that incorporated a total of 16 observational studies with 34 043 individuals and captured nearly 2600 major CVD events over a median follow-up of 6.9 years.7 This meta-analysis concluded that NAFLD (diagnosed by liver biopsy or imaging methods) conferred an OR of 1.64 for fatal and/or non-fatal CVD events (random-effects OR 1.64, 95% CI 1.26 to 2.13) (figure 1).7 Furthermore, risk of incident CVD events appeared to increase further with greater severity of NAFLD (random-effects OR 2.58; 95% CI 1.78 to 3.75) (figure 2), and remained statistically significant in those studies where analysis was fully adjusted for established CVD risk factors.7

Figure 1

Random-effects meta-analysis on the risk of incident CVD events (fatal, non-fatal or both) associated with NAFLD. Forest plot of comparison of patients with NAFLD versus those without NAFLD. Data are derived from Targher et al 7 (reproduced with permission). CVD, cardiovascular disease; NAFLD, non-alcoholic fatty liver disease.

Figure 2

Random-effects meta-analysis on the risk of fatal and non-fatal CVD events associated with more ‘severe’ NAFLD (defined either by presence of hepatic steatosis on imaging plus either increased serum gamma-glutamyltransferase levels or high NAFLD fibrosis score or high 18F-fluoro-deoxyglucose uptake on positron emission tomography, or by increasing fibrosis stage on histology). Data are derived from Targher et al 7 (reproduced with permission). CVD, cardiovascular disease; NAFLD, non-alcoholic fatty liver disease.

Other large studies recently showed that NAFLD was independently associated with an increased incidence of acute myocardial infarction, even in primary care populations.21 22 However, this latter finding has recently been questioned in a population-based case–control study that failed to find any significant association between a recorded diagnosis of NAFLD and risk of developing myocardial infarction and stroke, after adjustment for traditional CVD risk factors, using electronic records from four large European primary healthcare databases.23 However, the lack of any independent association between NAFLD and risk of acute myocardial infarction and stroke reported in this study23 may not be because such an association does not exist; but it is probably due to misclassification bias of NAFLD cases and other important methodological issues within the study design.24

It is worth noting that some observational cohort studies, mostly performed in Asian populations, have also reported that there is a significant and independent association between NAFLD and long-term risk of progression of subclinical coronary or carotid atherosclerosis, and, most importantly, that regression of NAFLD on ultrasonography over time is associated with a lower risk of carotid atherosclerosis development.25 26

Finally, convincing evidence indicates that NAFLD is strongly associated with valvular heart disease (mainly aortic valve sclerosis and mitral annulus calcification), increased risk of cardiomyopathy (mainly left ventricular dysfunction and hypertrophy, leading to the development of heart failure), arrhythmias (mainly permanent atrial fibrillation and increased QTc interval prolongation) and some cardiac conduction defects (mainly persistent first-degree atrioventricular block and left anterior hemiblock).27 28

Collectively, the available evidence not only demonstrates the strong association between NAFLD and CVD but also supports the view that NAFLD may increase risk of incident CVD events. These findings may have important implications for decision making in public health and clinical practice, and highlight the urgency of developing effective treatments for NAFLD. On this background of evidence, the European and American (published by the American Association for the Study for Liver Diseases (AASLD)) society guidelines for the management of NAFLD strongly recommended that all patients with NAFLD should undergo careful cardiovascular surveillance.29 30 To this end, a possible strategy at least in adults with NAFLD on primary CVD prevention might be to rely on the use of the Framingham risk score or other risk charts for CVD risk assessment.29–32 Although the Framingham risk score has been validated for use in NAFLD patients33 34, it remains to be demonstrated whether addition of NAFLD improves the accuracy of risk score systems to predict CVD events. Moreover, large randomised controlled trials (RCTs) with CVD outcomes that focus on treatments for liver disease in NAFLD are also needed to better establish a causal relationship between treatment of NAFLD and effects of improvements in liver disease on incident CVD events. Despite tremendous research advancements in NAFLD, our understanding of sex differences in NAFLD remains insufficient.35 It is known that CVD and NAFLD are both modulated by advancing age, sex, reproductive stage (ie, menopausal status) and synthetic hormone use.3 36–39 Recent evidence also suggests that women with NAFLD lose the CVD protection conferred by the female sex, and their global risk is underestimated by current CVD risk score systems.40 An adequate consideration of age, sex differences, sex hormones/menopausal status and other reproductive information in clinical investigation and gene association studies of NAFLD will be required to fill current gaps and implement precision medicine for NAFLD patients.35 In the meantime, also in accord with the AASLD clinical guidelines, we strongly recommend that aggressive modification of coexisting cardiometabolic risk factors should be considered in all patients with NAFLD as these patients are at high risk for CVD mortality and morbidity.30

### Mechanisms linking NAFLD to CVD and other cardiac complications

The pathophysiology behind the association of NAFLD with CVD and other cardiac complications is incompletely understood and may involve other pathways besides insulin resistance, for example, low-grade inflammation, oxidative stress and the effects of perturbations in the gut microbiota. (figure 3)41 Low-grade systemic inflammation is a key feature of many metabolic diseases, such as T2DM, obesity and related disorders including NAFLD. NAFLD is not only linked to CVD and T2DM, but also to chronic kidney disease.10 Importantly, these associations are especially relevant in patients with NASH, suggesting that liver inflammation may directly contribute to the development of these extrahepatic diseases.

Figure 3

Putative mechanisms linking NAFLD to ischaemic heart disease and other cardiac complications. Low-grade systemic inflammation plays a crucial role in the pathophysiology of cardiomyopathy and arrhythmias associated with NAFLD, and may also contribute to the development of ischaemic heart disease. In NAFLD, low-grade systemic inflammation is generated by complex inter-relationships between diet/food, the gastrointestinal tract, host factors such as genetics, the visceral adipose tissue and the liver. the liver is a major cytokine producer in NAFLD. NAFLD, non-alcoholic fatty liver disease; NASH, non-alcoholic steatohepatitis.

### Multiple sources of cytokines drive liver inflammation and extrahepatic complications

Whereas it is recognised that liver fibrosis determines long-term liver prognosis in NAFLD, it is generally accepted that liver inflammation precedes fibrosis in most instances. However, hepatic fat accumulation may also lead to liver damage, that is, fibrosis, independent from inflammation.42 In addition, it is well recognised that advanced disease (ie, fibrosis stage 3–4) is characterised by hepatic fat loss and less inflammation but increased adiponectin levels potentially contributing to this phenotype.43 Importantly, liver inflammation is accompanied by hepatic accumulation of inflammatory leukocytes and increased hepatic and extrahepatic cytokine production.44 45 It has also to be acknowledged that inflammation might be present in the liver intermittently and/or in a chronic-relapsing manner. This could also explain why liver fibrosis might play a role in CVD development.46 Many preclinical and clinical studies have shown that blockade of proinflammatory cytokines, such as interleukin (IL)-11, not only attenuates steatosis but also liver inflammation and fibrosis development47. Although various sites of cytokine production are assumed, such as the liver, adipose tissue and gastrointestinal tract, it remains unclear how much each compartment contributes to overall inflammation observed in NAFLD. The ‘multiple-hits’ hypothesis proposed a decade ago highlighted these different compartments as sources of cytokine production.44

Various lipid-related pathways may ‘drive’ hepatic inflammatory pathways in NAFLD.48 49 Whereas it had been initially believed that mainly intrahepatic triglyceride accumulation might contribute to liver inflammation, several studies have highlighted other pathways that may increase inflammation. These include enzymes involved in fatty acid synthesis, certain sphingolipids and polyunsaturated-derived eicosanoids, and specialised proresolving lipid mediators.50 Saturated fat induces more pronounced increases in intrahepatic triglyceride content and insulin resistance compared with unsaturated fat and simple sugars.51 Plasma lipids might also be disease relevant as shown for certain ceramides which concentrations were independently associated with greater severity of coronary artery stenosis in the left anterior descending artery.52 Mitochondrial dysfunction and endoplasmic reticulum stress activation are also key factors contributing to NAFLD and insulin resistance.53 Reducing endoplasmic reticulum stress by lipid chaperones reduces atherosclerosis, a key component in the clinical presentation of NAFLD.54

A link between dyslipidaemia and hepatic inflammation has also been suggested by recent data showing that proprotein convertase subtilisin/kexin type-7 gene variations correlate with severity of liver disease in human NAFLD.55 Furthermore, the presence of liver fat has also been linked to plasma inflammatory biomarkers in the Framingham Heart Study.56 Extracellular vesicles released by steatotic hepatocytes are also able to drive endothelial inflammation and atherogenesis.57 These vesicles are characterised by altered miRNA expression profiles facilitating vascular inflammation by miR-1 release and nuclear factor (NF)-κB activation.57 Besides the importance of pathways in adipose tissue, plasma lipids appear to be of crucial relevance in the association between NAFLD and CVD risk.58 Certain genetic variants associated with NAFLD, such as the patatin-like phospholipase domain-containing protein 3 (PNPLA3) and the transmembrane 6 superfamily 2 (TM6SF2) gene variants may protect against CVD risk and variants in glucokinase regulatory protein (GCKR) may be associated with increased CVD risk, perhaps mediated by a decrease in the atherogenic dyslipidaemic lipid profile in both PNPLA3 and TM6SF2 carriers and increase in the atherogenic dyslipidaemic profile in GCKR carriers.58 However, further research is needed to better understand whether ‘genetic-related NAFLD’ and ‘metabolic-related NAFLD’ may exert differential effects on risk of incident CVD events.10 59

Expanded visceral adipose tissue is a major site of low-grade systemic inflammation in NAFLD. Increased plasma IL-6 concentrations have also been associated with subclinical atherosclerosis in population-based studies60, and earlier studies have shown that visceral adipose tissue contributes at least 35% of circulating levels of IL-6, a major proinflammatory cytokine in obesity-related disorders that is mainly responsible for increased plasma C reactive protein levels.61 Visceral adipose tissue also expresses much higher concentrations of IL-6, IL-1β and tumour necrosis factor (TNF)-α compared with the liver and profound weight loss almost eliminates this expression, especially in adipose tissue.62 63 Expanded visceral adipose tissue might also affect NAFLD not only via the secretion of proinflammatory mediators as pharmacological inhibition of adipose triglyceride lipase by atglistatin inhibits high-fat diet induced insulin resistance and NAFLD64, establishing also a non-inflammatory ‘adipose tissue-liver’ axis.

### Proinflammatory pathways targeting vessels and the heart in NAFLD

Ectopic fat depots in the epicardium, pericardium and myocardium are associated with NAFLD and characterised by distinct metabolic signatures as demonstrated by magnetic resonance spectroscopy.65 To date, it is not known whether proinflammatory pathways in ectopic fat directly affect cardiac function and atherosclerosis development. In systemic inflammatory diseases such as rheumatoid arthritis, there is an increased risk of sudden cardiac death and arrhythmias66, and macrophage-derived IL-1β induces arrhythmias in diabetic mice.67 A meta-analysis has shown that increased proinflammatory biomarkers, such as plasma IL-6 and C reactive protein levels, are associated with an increased incidence of atrial fibrillation68, and whether decreasing pro-inflammatory biomarkers with a targeted anti-inflammatory agent reduces risk of CVD events has been tested in the Canakinumab Antiinflammatory Thrombosis Outcome Study (CANTOS) trial.69 In this proof-of-concept RCT, treatment with canakinumab (ie, an anti-IL-1β monoclonal antibody) led to a lower rate of recurrent CVD events than placebo, independent of lipid-level lowering.69 Also other proinflammatory cytokines, such as TNF-α and IL-17, have the capability, at least in preclinical studies, to induce cardiac arrhythmias.70 Intense physical endurance induced atrial arrhythmia susceptibility in rats, via a TNFα-dependent mechanism.71 IL-17, another proinflammatory cytokine, contributes to ischaemia-induced arrhythmias in rabbits72, and IL-1β, TNFα and IL-6 contribute to arrhythmias in rats.73

### NAFLD, microbiome and low-grade inflammation

There is an increasing evidence that the gut microbiota controls metabolic functions and is involved in NAFLD pathogenesis. Early animal experiments suggested that the gut microbiota is crucial for development of adipose tissue74 and evolution of NAFLD.75 A potential role for the intestinal microbiota in human NAFLD has been recently presented.76 Advanced liver fibrosis was associated with an increased abundance of Proteobacteria and Escherichia coli and a decrease in Firmicutes. Interestingly, a gut microbiome-specific signature has been demonstrated in NAFLD-related cirrhosis.77 Also in children an inflammation-related and fibrosis-related gut microbiome signature was observed with high presence of Prevotella copri. 78 At a species level, concentrations of Ruminococcus, Blautia and Dorea were increased in NASH patients.79 A profound intestinal dysbiosis has also been observed in NAFLD that is independent of obesity and insulin resistance.80 Faecalibacterium prausnitzii, a well-defined anti-inflammatory bacterial strain, was substantially decreased in NAFLD patients81 and substantial changes in the gut microbiome with a decrease in Collinsella and Parabacteroides have been observed in NAFLD-associated coronary heart disease.82

The gut microbiota affects substantially circulating metabolites in NAFLD.83 Phenylacetate is associated with hepatic steatosis and faecal transfer from obese women with high-grade steatosis into mice resulted in hepatic steatosis, as did feeding phenylacetate to mice.83 Other gut-derived metabolites might be involved in NAFLD pathogenesis.84 3-(4-hydroxyphenyl)-lactate, mainly derived from Proteobacteria, was associated (in two independent patient cohorts) with hepatic steatosis and degree of fibrosis.84 Bacterial components may also be present in the livers in NAFLD, as a meta-taxonomic signature and also increased endotoxin has been detected in the livers.85 86. All these studies support a role for intestinal microbiota in NAFLD pathogenesis and hold the promise that manipulation at this level might improve liver disease phenotype. That said, to date, it remains uncertain what prebiotics, probiotics or synbiotics should be used to change the gut microbiota. Moreover, it is not known which gut microbiota need to be modified, both in type and in quantity, in order to benefit the liver and/or CVD risk in NAFLD. A recent phase 2 RCT tested whether 1-year administration of a synbiotic combination of probiotic and prebiotic agents affected hepatic fat content (assessed by magnetic resonance spectroscopy), non-invasive fibrosis biomarkers, and the composition of the faecal microbiome in 104 UK patients with NAFLD. The results of this RCT showed that the synbiotic altered the faecal microbiome, but did not reduce hepatic fat content or biomarkers of liver fibrosis. Faecal samples from patients, who received the synbiotic, had higher proportions of Bifidobacterium and Faecalibacterium, and reductions in Oscillibacter and Alistipes, compared with baseline (changes were not observed in the placebo group).87

### Trimethylamine N-oxide: prototypic microbiota-derived metabolite contributing to CVD

Gut microbes and related metabolites have been recently discovered as potentially important players in CVD. Commensals convert certain nutrients such as choline or carnitine into trimethylamine (TMA), which is metabolised in the liver by flavin mono-oxygenases to TMA N-oxide (TMAO). L-carnitine enriched diet in humans is converted into TMAO, an effect which was less pronounced in vegans/vegetarians.88 This has also been observed after chronic red meat consumption and, interestingly, discontinuation of red meat consumption reduced TMAO levels within 4 weeks.89

Many studies have shown that higher circulating TMAO levels are associated with adverse CVD outcomes.90 91 Furthermore, some meta-analyses have also confirmed a strong association between circulating TMAO levels and risk of fatal and non-fatal CVD events.92 93 Patients with ischaemic stroke also exhibit higher TMAO levels than healthy controls94, and increased TMAO levels also predict CVD mortality in patients with existing peripheral arterial disease.95 Interestingly, so far only a very few reports exist investigating circulating TMAO levels in NAFLD cohorts. In a study assessing 60 biopsy-proven NAFLD subjects, a greater severity of NAFLD was associated with higher TMAO levels but lower betaine and betaine/choline ratio.96 Despite the shortage of reports on TMAO in NAFLD, it is increasingly accepted that circulating TMAO levels are a prominent biomarker of CVD, which is the most common cause of mortality in NAFLD patients. In addition, an association with thrombosis events was both shown clinically and experimentally as TMAO alters calcium signalling in platelets, and enhances responsiveness and in vivo thrombosis potential in various animal models.97. Inhibitors of TMA-generating enzymes significantly reduced plasma TMAO levels for up to 3 days and rescued diet-induced enhanced platelet responsiveness and thrombus formation.98 Another study observed a U-shaped association between TMAO levels and mortality risk in patients with acute venous thromboembolism, but it was not associated with recurrent venous thromboembolism.99

The explanation as to how elevated TMAO levels might increase risk of CVD/cardiac complications is uncertain. A recent study found that TMAO affects the cardiac autonomic nervous system, promoting ischaemia-induced ventricular arrhythmias.100 Another mode of action might involve the endoplasmic reticulum stress kinase PERK (ie protein kinase RNA-like endoplasmic reticulum kinase). PERK is a receptor for TMAO, and its binding results in PERK activation and induction of the transcription factor FoxO1, a key factor in metabolic disorders.101 Interestingly, TMAO may directly activate pro-inflammatory pathways as it upregulates NLRP3 and NF-κB and thereby promotes vascular calcification.102 Thus, TMAO reflects a crucial microbiota-derived biomarker of atherosclerosis and potentially of NAFLD-associated CVD.

### Pharmacological treatment

The cornerstone of NAFLD management remains lifestyle modification. Weight loss, increased physical activity and reductions in coexisting cardiometabolic risk factors may all have beneficial effects in NAFLD. Weight loss of approximately 5%–7% is able to decrease hepatic steatosis; however, an approximate 10% wt loss is required to reverse NASH and weight loss of ≥10% may also improve or reverse hepatic fibrosis.29 30 Additionally, bariatric or weight loss surgery has been shown to ameliorate many CVD risk factors and may also be directly beneficial in patients with early liver disease. However, it is beyond the remit of this review to discuss the metabolic and vascular benefits of bariatric surgery in NAFLD and the reader is referred to recent clinical practice guidelines for the perioperative nutrition, metabolic and non-surgical support of patients undergoing bariatric procedures.103

Presently, there are no approved pharmacological treatments for NAFLD or NASH. From a recent systematic review, it clearly emerges that the major issue in this field is the scarcity of high quality, adequately powered RCTs of sufficient duration that include clinically relevant hepatic endpoints (ie, liver histological data).104 However, there are several novel therapeutic agents under active investigation, and a variety of other drugs will also likely emerge over the next few years, allowing a more staged approach to the management of NAFLD that is likely to vary from patient to patient. That said, in selecting a specific drug for the treatment of NAFLD, we believe that pharmacological treatments should be chosen that target not only liver-related complications (cirrhosis and hepatocellular carcinoma (HCC)) but also the increased CVD risk in NAFLD.105 Additionally, since NAFLD is a risk factor for incident T2DM106 (which is also a risk factor for CVD), the ideal treatment for NAFLD would not only ameliorate liver disease, but also attenuate risk of developing T2DM107, and thereby consequently lessen the risk of CVD.

It is beyond the scope of this review to discuss the evidence for all drugs that have been tested in the treatment of NAFLD. Therefore, we have focused on drug treatment options that might benefit not only the liver but also have beneficial (or adverse effects) on NAFLD-associated CVD risk. As discussed above, there is also a growing interest in the role of dysbiosis in both the pathogenesis of NAFLD and CVD. Whether faecal transplantation108 to improve the gut microbiota profile and drugs relevant to the treatment of NASH, can favourably affect: gut microbiota; modify intestinal permeability and intestinal functions; and thereby treat NAFLD and CVD, remains uncertain. Presently, it is not known whether faecal transplantion benefits NAFLD. However, a recent pilot in which 20 men with metabolic syndrome were randomised to single lean vegan-donor or autologous faecal microbiota transplantation, caused detectable changes in intestinal microbiota composition, but failed to induce changes in TMAO production capacity or parameters related to vascular inflammation.109

We have also briefly discussed below the evidence to date showing whether (or not) drugs relevant to the treatment of NAFLD and CVD can affect the gut microbiota, or gut microbiota-related mechanisms relevant to liver and vascular diseases.

### Pioglitazone

The discovery of peroxisome proliferator-activated receptor gamma (PPAR-γ) in adipose tissue produced a step change in adipose tissue research.110 PPARs are a group of nuclear receptor proteins that function as transcription regulators and PPAR-γ heterodimerises with retinoid X receptor and binds to specific DNA sequences to regulate adipocyte differentiation and function, lipid metabolism and inflammation.111 Glitazones (eg, rosiglitazone and pioglitazone) are selective activators of PPAR-γ and pioglitazone is a potent insulin sensitizer that is currently licensed for treatment of T2DM. Although there are well-recognised side effects of pioglitazone, such as a mild increase in body weight (especially subcutaneous fat depots), fluid retention (oedema and heart failure) and an increase in fragility fractures, there are also many benefits of pioglitazone besides its very durable effect to reduce plasma glucose concentrations in people with T2DM.

Since NAFLD independently increases risk of incident T2DM by ~2.2-fold106 and pioglitazone decreases risk of incident T2DM in individuals with pre-diabetes112, it is reasonable to assume that pioglitazone may also decrease risk of incident T2DM in patients with NAFLD. Moreover, NAFLD increases risk of hypertension113, a recognised CVD risk factor, and pioglitazone lowers blood pressure.112 NAFLD is an independent risk factor for CVD7 and both ischaemic heart disease and stroke are two of the leading causes of death worldwide. T2DM also increases risk of major CVD events ~twofold114–116 and pioglitazone has been shown in the PROactive trial (PROspective pioglitAzone Clinical Trial In macroVascular Events) to decrease the composite of all-cause mortality, non-fatal myocardial infarction or stroke in T2DM patients with macrovascular disease.117 In this RCT, pioglitazone use was also associated with a 28% decrease in myocardial infarction118 and a 47% decrease in ischaemic stroke.119 In support of these findings, a meta-analysis of 19 RCTs enrolling ~16 500 patients showed a summary estimate of an 18% decrease in the composite of all-cause mortality, myocardial infarction or stroke (HR (HR) 0.82; 95% CI 0.72 to 0.94) with pioglitazone treatment.120 Another meta-analysis investigating the effect of pioglitazone on risk of CVD events showed a benefit with pioglitazone in both patients with pre-diabetes (or insulin resistance) and those with T2DM.121 Recent evidence also showed that pioglitazone decreased risk of stroke or myocardial infarction in patients without T2DM but with insulin resistance after previous stroke or transient ischaemic attack.122 123 A large umbrella review recently confirmed that pioglitazone significantly decreased risk of major CVD events but increased risk of heart failure.124

Pioglitazone treatment has been tested in several placebo-controlled RCTs in patients with biopsy-confirmed NASH and pioglitazone treatment resulted in improvement in histologic features of NAFLD and resolution of NASH in ~50% of patients; regardless of diabetes status.125–128 Interestingly, a recent meta-analysis of eight RCTs (including a total of 516 adults with biopsy-confirmed NASH) showed that pioglitazone improved advanced fibrosis in NASH, even in patients without diabetes.129 Although the PPAR-γ2 isoform is highly expressed in adipocytes, PPAR- γ1 isoform is also expressed in both hepatic stellate cells and Kupffer cells. Pioglitazone effects on the liver are likely mediated by a combination of indirect effects on the adipose tissue to decrease free fatty acid flux to the liver and increase adiponectin levels (resulting in improved hepatic steatosis); and a direct effect of the drug on both Kupffer cells and stellate cells to decrease hepatic inflammation and fibrogenesis. Based on the available evidence, three sets of guidelines from the UK, Europe and USA have strongly recommended pioglitazone for treatment of NASH.29 30 130

Although presently it is not possible to predict which patients with NASH are going to achieve NASH resolution with pioglitazone use, a recent post hoc analysis of the Pioglitazone vs Vitamin E vs Placebo for Treatment of Non-Diabetic Patients with Nonalcoholic Steatohepatitis (PIVENS) trial126 suggested that after treatment with pioglitazone, patients with histological resolution of NASH had favourable changes in plasma lipoprotein subfractions compared with those without NASH resolution. In fact, individuals with NASH resolution had a significantly increased mean peak low-density lipoprotein (LDL) diameter and a higher frequency of LDL phenotype A (ie, large buoyant LDL particles) at week 96, even after adjustment for relevant confounding factors, including treatment group.131

To date, there is limited data regarding whether pioglitazone use may affect the gut microbiota. However, the PPAR-γ receptor is a butyrate sensor in the colonic lumen132, and microbiota-activated PPAR-γ signalling has been reported to prevent dysbiotic expansion of pathogenic bacteria by driving the energy metabolism of colonic epithelial cells.133 In a mouse model of dietary fructose-driven gut dysbiosis that caused intestinal epithelial barrier impairment134, the authors showed that pioglitazone repaired intestinal epithelial barrier damage by activating the NOD-like receptor family pyrin domain-containing 6 (NLRP6) inflammasome. Thus, it is possible that pioglitazone use could decrease the inflammatory stimulus from lipopolysaccaride breaching the intestinal epithelial barrier, and gaining access to the portal circulation.

Such is the wealth of evidence supporting its effectiveness in decreasing risk of incident T2DM, treating hyperglycaemia in T2DM and decreasing risk of major CVD events, pioglitazone has been recently described as the ‘forgotten, cost-effective, cardio-protective’ drug for T2DM.135 Given the evidence described above supporting its use in the treatment of liver disease in NASH, the overall evidence supports its use in NASH assuming there are no contradictions to treatment with pioglitazone. Few drugs are free of side effects and clinicians need to weigh up the balance of risk and benefits of prescribing this drug in their individual patients with NASH. Figure 4 schematically shows the inter-relationships between NAFLD, T2DM and CVD and where RCTs have shown pioglitazone treatment acts to significantly decrease risk of clinical outcomes in each condition. Were it not for the fact that pioglitazone treatment is associated with an increased risk of weight gain, and a small increase in bone fracture risk, pioglitazone treatment would be much more widely used in treating patients with NASH.

Figure 4

The figure schematically summarises the inter-relationships between each condition (ie, NAFLD, T2DM and CVD) from the results of prospective cohort studies and also illustrates where randomised controlled trials have shown pioglitazone treatment acts to decrease risk of clinical outcomes. CVD, cardiovascular disease; Mets, metabolic syndrome; MI, myocardial infarction; NAFLD, non-alcoholic fatty liver disease; NASH, non-alcoholic steatohepatitis; T2DM, type 2 diabetes mellitus.

### Statins

There is limited high-quality data with histological liver endpoints showing that statin use improves NASH.136 There is also limited data regarding whether statin use affects the gut microbiota. That said, it has been suggested that the modulation of gut microbiota by statins has an important role in the therapeutic actions of these drugs137, and these authors also suggested that faecal microbiota transplantation also improved plasma glucose concentrations. In this study using a mouse model of high-fat diet-induced obesity, the association between gut microbiota and immune responses was investigated. Both atorvastatin and rosuvastatin increased the abundance of the genera Bacteroides, Butyricimonas and Mucispirillum. The abundance of these genera was correlated with the inflammatory response, including levels of IL-1β and transforming growth factor-β1 in the ileum. In addition, oral faecal microbiota transplantation with faecal material collected from rosuvastatin-treated mouse groups improved hyperglycaemia. Additionally, a proof-of-concept study in individuals with dyslipidaemia showed that 4–8 weeks of rosuvastatin treatment significantly altered the gut microbiome and the abundance of specific bacterial taxa, which was correlated with the LDL-cholesterol-lowering response of the drug.138 In this study, both Firmicutes and Fusobacteria were inversely associated with plasma LDL-cholesterol concentrations, while Cyanobacteria and Lentisphaerae were positively associated with LDL-cholesterol concentrations. However, it is important to note that this study lacked a control group, and the bacterial sequencing was performed only after rosuvastatin treatment. Consequently, the authors did not investigate the changes in the gut microbiome. Finally, it has also been suggested that gut microbiota may interact with statin treatments to both modify farnesoid X receptor signalling and decrease statin bioavailability, thereby potentially producing physiologically relevant effects on liver lipid and glucose metabolism.139

A recent Expert Panel Statement concluded that the evidence from: animal studies, five post hoc analyses of prospective long-term survival studies, and five rather small biopsy-proven NASH studies that investigated the effect of statins on the liver in NAFLD, was not good enough to recommend statin treatment specifically for treating liver disease in NAFLD.140 Notably, these studies provided data that suggested biochemical and histological improvement of NAFLD/NASH with statins and, in the clinical studies, large reductions in CVD events in patients with NAFLD compared with those who did not have NAFLD.140 Recently, there has also been interest in whether statins specifically decrease risk of liver fibrosis. In a cross-sectional study of 346 individuals with T2DM of which 45% were taking statins, multivariate analyses showed that statin use was inversely associated with significant liver fibrosis, despite statin-treated patients being older, more frequently male and with poorer glycaemic control than those without statins.141 However, it should be noted that to date, none of the available evidence is from RCTs that have tested the prior hypothesis that statins decrease liver fibrosis. Thus, the evidence is currently not good enough to recommend statin usage in order to specifically treat NAFLD or NASH. Nevertheless, pending forthcoming RCTs, clinicians should consider combining statins and pioglitazone in those patients with NAFLD or NASH, who are at high risk of CVD, for the primary and secondary prevention of CVD.140

Currently, the American College of Cardiology/American Heart Association guidelines for primary CVD prevention recommend statin use as a first-line treatment in patients with increased plasma LDL-cholesterol concentrations (LDL-cholesterol ≥5 mmol/L); those with T2DM, who are 40–75 years of age; and those determined to be at ‘sufficient’ CVD risk.142 Presently, there is disagreement between different professional societies as to what constitutes ‘sufficient’ CVD risk (to prescribe statins), but in the above guidelines ‘sufficient’ CVD risk is defined as ≥7.5% risk of developing a CVD event over 10 years. Although the CVD risk threshold that is required to advocate statin treatment has been lowered considerably over the last 20 years, most professional societies would endorse statin treatment when the patient’s 10 year CVD risk estimation was ≥10%.

For estimating CVD risk in NAFLD patients, there are no specific CVD risk prediction tools that take into account the presence or severity of NAFLD. To date, there is insufficient evidence to gauge whether knowing the patient has a diagnosis of NAFLD (with or without accompanying fibrosis) adds to existing risk factors in CVD risk estimation. Consequently, rather than recommending any specific CVD risk calculator, for example, the Framingham risk score or the SCORE (Systematic Coronary Risk Estimation) charts, it is better that a clinician uses a risk calculator than not. Given that the evidence discussed above suggests that NAFLD is a risk factor for CVD, it is highly likely that prediction of 10-year CVD risk in NAFLD is an underestimate of true CVD risk. Consequently, since statins are safe in patients with NAFLD143, it would seem logical to err on the side of caution, and advocate use of statins to decrease CVD risk when the 10-year CVD risk is ≥7.5%. There is also some, more limited, evidence that statin treatment is associated with a reduced risk of HCC, most strongly in Asian but also in Western populations.144 However, RCTs with statin treatment are required in populations at high risk of HCC, before advocating this treatment specifically to attenuate risk of HCC.

### Metformin and other newer antihyperglycaemic agents

Metformin represents the first-line choice for treatment of T2DM worldwide. However, metformin is not currently recommended as a specific treatment for NAFLD or NASH, mostly due to its lack of efficacy on hepatic histological endpoints in both adults and adolescents with biopsy-confirmed NASH, irrespective of diabetes status.29 30 104 To date, there remains uncertainty about whether metformin reduces risk of major CVD events.124 145 Interestingly, however, several preclinical and observational studies and recent meta-analyses suggest that metformin reduces risk of developing some types of cancer, especially HCC.146 147 It has also become well accepted that metformin has favourable effects on the intestinal microbiome. Metformin treatment increases microbial diversity and specifically increases mucin-degrading Akkermansia muciniphila, as well as several short-chain fatty acid-producing microbiota, increasing levels of butyrate and propionate that are involved in both glucose homoeostasis and maintaining colonic epithelial integrity.148 149

Similar to metformin, no robust RCT data exist with histological liver endpoints as a primary outcome to formally comment on the effectiveness of the use of the newer antihyperglycaemic agents, such as dipeptidyl peptidase-4 (DPP-4) inhibitors, glucagon-like peptide-1 receptor agonists (GLP-1 RAs) or sodium glucose cotransporter 2 (SGLT-2) inhibitors as a treatment for NAFLD or NASH29 30 104, as shown in table 1. Among these newer anti-hyperglycaemic drugs, GLP-1 RAs seem to exert the most promising beneficial effects on NAFLD or NASH. A recent systematic review examining the efficacy of antihyperglycaemic drugs in patients with biopsy-proven or imaging-defined NAFLD with or without T2DM has supported the capability of GLP-1 RAs to reduce serum liver enzyme levels and improve NAFLD as detected by imaging techniques or liver histology.104 In particular, a phase 2 RCT involving 55 UK obese patients with biopsy-proven NASH, it has been shown that patients who were randomly assigned to liraglutide 1.8 mg/day for 48 weeks had a greater histological resolution of NASH and significant improvements in individual histologic scores of NASH compared with those receiving placebo.150 The authors suggested that the beneficial effects of liraglutide on the histological liver endpoints were due both to its direct hepatic effect and to concomitant weight loss as liraglutide is a potent treatment to effect weight loss.150 Importantly, liraglutide and other long-acting GLP-1 RAs have also been shown to reduce risk of adverse CVD and renal outcomes in patients with T2DM.124 151 For such reasons, if larger phase 3 RCTs confirm the promising findings of this RCT, it would be reasonable to assume that GLP-1RAs will become a treatment option in NASH, especially in those patients who are obese or have T2DM. A recent comparison of the effects of treatment with metformin vs the GLP-1 agonist liraglutide on the gut microbiota in patients with T2DM showed that patients taking metformin had a significant increase in the relative abundance of the bacterial genus Sutterella, whereas those taking liraglutide had a significant increase in the genus Akkermansia. Thus, these preliminary data suggest that these two anti-hyperglycaemic drugs have differential effects on the microbiome, despite the fact that both drugs are similarly effective in lowering plasma glucose concentrations.152

Table 1

Principal phase two placebo-controlled or head-to-head RCTs testing the efficacy and safety of antihyperglycaemic drugs in patients with NAFLD or NASH

A systematic review also supported the possibility that SGLT-2 inhibitors may improve liver fat content (as assessed by imaging techniques) and serum liver enzymes.104 However, most of the RCTs testing these novel drugs are small with a short period of follow-up, and importantly, to date, there are no placebo-controlled RCTs examining the long-term effects of SGLT-2 inhibitors on histologic features of NAFLD.104 Additionally, there is also very limited data as to the effects of this class of drugs on the gut microbiome. SGLT-2 inhibitors have been shown to consistently reduce risk of major CVD events, heart failure and renal outcomes in patients with T2DM.124 153 Moreover, among patients with systolic heart failure, the risk of worsening heart failure or of CVD mortality was lower among those patients who received the SGLT-2 inhibitor dapagliflozin than among those who received placebo; regardless of the presence or absence of T2DM.154 Thus, this effect may represent an attractive bonus for the use of SGLT-2 inhibitors in NAFLD.

### Obeticholic acid and other drugs

A number of phase 2 and phase 3 head-to-head or placebo-controlled RCTs have tested the efficacy and safety of novel drug treatments in NAFLD or NASH (table 2). Of these, obeticholic acid is one of the more promising new agents for NASH treatment. Obeticholic acid is a selective farnesoid X receptor agonist that regulates bile acid and lipid metabolism. Obeticholic acid at a dose of 25 mg/day has effected significant improvements in liver histology in the phase 2 FLINT (Farnesoid X Receptor Ligand Obeticholic Acid in NASH Treatment) trial155, as well as well as positive ad interim results in the ongoing phase 3 REGENERATE trial.156 Obeticholic acid was also associated with a mild decrease in body weight. However, in both trials, obeticholic acid caused marked increases in plasma LDL-cholesterol levels (nearly a 40 mg/dL increase) within 1 month of treatment (and more than half of patients treated with obeticholic acid started statin therapy in the REGENERATE trial).155 156 Recently, it has been suggested that obeticholic may also modify the gut microbiota and produce a favourable effect on the gut microbiome.157 In this experimental study, treatment with antibiotic (that removed normal commensal bacteria) attenuated the effect of obeticholic acid in mice. Obeticholic acid treatment markedly increased abundance of Blautia and the concentration of taurine-bound bile acid induced by the high fat diet was reduced in liver.157 In a phase 1 RCT in man, treatment with obeticholic acid for 17 days, that suppressed bile acid synthesis, produced a reversible induction of Gram-positive bacteria that are found in the small intestine. There was also an increase in the representation of microbial genomic pathways involved in DNA synthesis and amino acid metabolism with obeticholic acid treatment.158

Table 2

Principal phase 2 or phase 3 head-to-head placebo-controlled RCTs (published in the last 10 years) testing the efficacy and safety of non-antihyperglycaemic drug treatments on NAFLD or NASH (assessed either by liver biopsy or by MRI) in overweight or obese adult individuals

### Supplemental material

In a two phase RCT, a 1-year treatment with elafibranor 120 mg/day (ie, a dual agonist of peroxisome-proliferator activated receptor [PPAR]-α and PPAR-δ) was significantly associated with a higher rate of NASH resolution than occurred in the placebo arm. Elafibranor also improved plasma LDL-cholesterol, triglyceride and glucose levels.159 It is uncertain whether elafibranor modifies the gut microbiome, and longer-term phase 3 RCTs are also required to confirm the positive effects of elafibranor on the liver in NASH.

## Conclusions

This review supports the notion that CVD is the leading cause of death in NAFLD patients and that NAFLD is closely associated with an increased risk of CVD events and other cardiac complications (ie, cardiomyopathy, cardiac valvular calcification and arrhythmias) independent of traditional cardiovascular risk factors and metabolic syndrome features. Although further research is needed to draw a definitive conclusion, these observations raise the possibility that NAFLD, especially its more advanced forms, is directly involved in the pathogenesis of CVD. Recent evidence discussed here suggests that this process is mediated not only via the atherogenic dyslipidaemia occurring with features of the metabolic syndrome and NAFLD, but also through the systemic release of multiple proinflammatory and proatherogenic mediators from both the steatotic and inflamed/fibrotic liver and the intestine via changes in gut microbiota. The existing evidence to date reinforces the notion that NAFLD is a multisystem disease affecting many extrahepatic organ systems, including the cardiovascular system. Thus, we believe that a purely ‘liver-centric’ approach to NAFLD is not sufficient and treatment of this burdensome liver disease needs to shift to a more patient-centred, multidisciplinary team-based approach. Since more patients with NAFLD will die from CVD than from the consequences of their liver disease, we strongly believe that a careful assessment of the 10-year CVD risk is mandatory in all persons with NAFLD, together with early and aggressive treatment of all coexisting cardiometabolic risk factors.

## Acknowledgments

CDB is supported in part by grants from the Southampton National Institute for Health Research Biomedical Research Centre. GT is supported in part by grants from the University School of Medicine of Verona, Verona, Italy. HT is supported by the excellence initiative VASCage (Centre for Promoting Vascular Health in the Ageing Community), an R&D K-Centre (COMET program - Competence Centers for Excellent Technologies) funded by the Austrian Ministry for Transport, Innovation and Technology, the Austrian Ministry for Digital and Economic Affairs and the federal states Tyrol, Salzburg and Vienna.

## Footnotes

• GT, CDB and HT contributed equally.

• Correction notice This article has been corrected since it published Online First. The ORCID ID's have been added and table 2 has been corrected.

• Contributors All authors contributed equally to write the manuscript.

• Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

• Competing interests None declared.

• Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

• Patient consent for publication Not required.

• Provenance and peer review Commissioned; externally peer reviewed.